Polygenic Risk Scores for Alzheimer's Disease in Hispanic/Latinx Populations

点击翻译按钮获取中文摘要

ABSTRACT. Late-onset Alzheimer's disease (LOAD) is the most common neurodegenerative disease, but its causes remain largely unclear. LOAD's large missing heritability has been attributed, at least partially, to mechanisms not currently investigated by traditional “single-locus” approaches. To this end, we aim to study the role of Polygenic risk scores (PRS), which capture sparse genetic information by summing up multiple risk loci across the genome, providing an individual genetic risk profile. Numerous studies have confirmed that LOAD is enriched with a polygenic component. Most of PRS have been developed in European-descend populations and no ancestry-specific PRS for LOAD have been developed in Hispanic/Latinx (HL). Furthermore, most PRS rely on common genetic variants (minor allele frequency ≥1%), neglecting the contribution of rare variants which have been shown to be critical in LOAD pathogenesis. To this end, we will generate traditional and innovative PRS leveraging large HL cohorts available at Columbia University, employing common and rare variants. Because linkage disequilibrium between common and rare variants is low, we hypothesize that these two PRS will be independent in terms of disease prediction and patient risk stratification. For common variants PRS (cvPRS), we have strong preliminary data in Caribbean Hispanics that culminated in a recent publication from our group. In addition, we will employ a new approach developed by our group, RV- EXCALIBER, which combines rare variant burden over a large number of genes into predictive rare variant polygenic risk score (rvPRS). Employing HL is particularly indicated for rvGRS because 1) they are enriched with rare variants and 2) have higher incidence and prevalence of LOAD compared to European descend populations. We will study three HL populations that exhibit different proportions of European, African and Amerindian ancestry: 1) Caribbean Hispanics 2) Mexicans and 3) Peruvians. These cohorts are deeply phenotyped for LOAD, have GWAS data and whole-exome/genome sequencing data (WGS) and plasma AD-biomarkers available for a sub-sample. We will first (Aim 1) generate a cvPRSs using common variants associated with LOAD within each Hispanic cohort. We will also generate an alternative local ancestry-weighted PRS and test the transferability of each cvPRS across HL cohorts and finally will construct a trans-ancestry PRS. Then (Aim 2) we will generate rvPRSs for those individuals with WGS and compare the performances of the two PRSs. We will also test a model that combine cvPRS and rvPRS. Finally, we will validate these new PRSs from Aim 1 and Aim 2 employing LOAD-related endophenotypes and explore a multi-trait PRS approach (Aim 3).